Everyone is writing about the new WPA3 Wi-Fi security standard, and how it improves security over the current WPA2 standard.

This summary is as good as any other:

The first big new feature in WPA3 is protection against offline, password-guessing attacks. This is where an attacker captures data from your Wi-Fi stream, brings it back to a private computer, and guesses passwords over and over again until they find a match. With WPA3, attackers are only supposed to be able to make a single guess against that offline data before it becomes useless; they'll instead have to interact with the live Wi-Fi device every time they want to make a guess. (And that's harder since they need to be physically present, and devices can be set up to protect against repeat guesses.)

WPA3's other major addition, as highlighted by the Alliance, is forward secrecy. This is a privacy feature that prevents older data from being compromised by a later attack. So if an attacker captures an encrypted Wi-Fi transmission, then cracks the password, they still won't be able to read the older data -- they'd only be able to see new information currently flowing over the network.

Note that we're just getting the new standard this week. Actual devices that implement the standard are still months away.

Posted on July 12, 2018 at 6:11 AM • 25 Comments


Aaron KelleyJuly 12, 2018 7:00 AM

Any chance of existing devices being updated to support WPA3, via OS/driver/firmware updates? I know that many vendors would not bother to make the attempt, but some including the leading "mesh network" vendors roll out regular firmware updates with new features from time to time. It seems like it shouldn't actually require new hardware in order to get up and running, but rather it could be implemented mostly in software?

M@July 12, 2018 7:31 AM

@Aaron Some of the recent gen gear will definitely update over firmware. I don't want to plug any given vendor here, but there's a good handful who are already cranking away on it. The same gear that is repeatedly compromised because no one ever updates it ever, however, will likely require hard changes, so get ready for your favorite blue-and-black-box vendors to try to sell you new hardware because they don't invest in better software delivery.

stdpunditJuly 12, 2018 8:15 AM

a large part of the WPA3 standard is the straight implementation of features present in the 2012 edition of the 802.11 Standard. Most notably, the first handshake is called SAE and has been introduced in 802.11s the year before. (The second handshake is ye good old KRACKed 4 way handshake).

wpa_supplicant and hostapd already support those things (which is logical, considering the maintainer contributes to the wi-fi and 802.11 standard), but your driver probably do not.

PeteJuly 12, 2018 8:24 AM

Based on the history for all other standards, I'll still be using a VPN even to connect to my home LAN from a wifi router I manage.

People are entirely too trusting of RF solutions. Add a VPN layer if you actually care about security.

ConfusedJuly 12, 2018 10:08 AM

"With WPA3, attackers are only supposed to be able to make a single guess against that offline data before it becomes useless"

Can someone please explain to me how that works?

RealFakeNewsJuly 12, 2018 11:26 AM

From what I've heard about WPA3: it's nothing special, and as written above, is a missed opportunity.

The whole thing needs re-working from scratch, and needs to get rid of all the junk that's included.

WardriverJuly 12, 2018 2:32 PM

What @confused said.

There isn't any way to do what they are advertising they can done regarding one-time password attempts without (a) making it impossible for the average user to use or (b) overstating how hard it is to crack.

"they'll instead have to interact with the live Wi-Fi device every time they want to make a guess. (And that's harder since they need to be physically present, and devices can be set up to protect against repeat guesses.)"

(a) Wait, what? I don't know on anyone who cracks wifi passowrds without being physically present to the wifi stream. The average range of a home router in perfect conditions is about 100 yards. People are not capturing a signal, flying to another country, and then cracking it. Don't be damn silly. They are sitting in their car with tinted windows down the block from your house...

(b) there is no way to limit repeated guesses for an attacker that don't also limit repeated guesses for the legitimate end user. This sounds like a money-making scheme for the router industry to force people to throw their routers away every time they forget their password.

@refalfakenews "TL;DR: most of the new features and improvments are optional so most vendors will do the bare minum and place "wifi certified" over it."

Of course they will be optional because the standard is security theater run amok.

WardriverJuly 12, 2018 2:50 PM


Here is the ieee paper about stopping off-line contradictory attacks. Honestly, I not enough of a crypto-geek to know if what they are talking about makes sense. But as an old wardriver I can't see it interfering with the work. It seems to be intent of defeating a class of attacks that by and large are just not done (except maybe by script kiddies who have just found out about Kismet and Jack The Ripper.)

MarkJuly 12, 2018 10:45 PM

The biggest change since ISPs started to ship routers with WPA2 enabled?

Honestly I view wireless security (WPA2 with a strong passphrase) as a solved problem, especially in corporate environments with user+certificate authentication.

Ross SniderJuly 13, 2018 1:49 AM

Is the summary here that popular editorials have glossed over the controversial decisions of an NSA-backed standard?

The use of the dragonfly PAKE has significant disadvantages:
1. There are known timing side channels on the protocol which already defeat the password protection.
2. There are several known active attacks on the protocol which already defeat the password protection, some under the assumption of non-robust implementations and others unconditional properties of the protocol.
3. The implication of the protocol is that passwords must be stored in a format which make them brute forceable, again defeating the password protection.
4. The parameter negotiation can be used to force malicious parameters, which is even more dangerous given how the protocol can be initiated by any side.
5. The implementation of the crypto is fragile (easy to exclude checks which subvert the security of the scheme). It is not clear that the specification includes all checks necessary for a robust implementation.

This is on top of there being no formal security proofs and a lack of crypt-analysis, while there are other already standardized alternatives. To note that these protocols are typically very difficult to design.

I question the IETF's decision, believe this standard to contain significant cryptographic weaknesses that probably amount to an NSA backdoor, and will attempt to use VPN over WPA3 wifi connections.

Some GuyJuly 13, 2018 2:18 AM

@swr: Dragonfly caused some controversy in the IETF:

It's not the protocol I'd have chosen. The author is a difficult person to work with and has a long history of trying to push his pet crypto projects through as standards despite multiple cryptographers pointing out problems with them, which he invariably perceives as some sort of personal attack on him rather than the genuine technical criticism which they are. It'll be interesting to see whether further attacks on Dragonfly turn up in the future (I would say, yes, they will). It'd also be interesting to find out what sort of shenanigans went on for Dragonfly of all things to get adopted for WPA3 when there are so many other well-designed, heavily-analysed protocols around.

WeatherJuly 13, 2018 2:40 AM

Trying to understand this new standard, the router sends data to the client which then uses that to work out the key to use. Is that correct can't you just bruteforce the data that the router sends that is used..
is the key worthless because it decrement a value making the key no longer valued??

RealFakeNewsJuly 13, 2018 6:04 AM

@Ross Snider: I thought that was the case but I wasn't sure I was confusing it with something else.

I think it's been debated here quite extensively in the Squid posts.

meJuly 13, 2018 6:31 AM

(a) Wait, what? I don't know on anyone who cracks wifi passowrds without being physically present...People are not capturing a signal, flying to another country, and then cracking it.

You are wrong, that's exactly what happens:
1-you go in range
2-you launch a deauthentication attack
3-you record the "login" handshake
4-you have everything you need, you can go offline or wherever you want.
5-you use one or more computers to try multiple passwords against the recorded handshake until you find one that match.
and people use *amazon cloud computing* to do that, or a desktop pc with multiple gpu.
they use it because is a super computer, quite faster than home pc.

in the new way you can't record "something", copy that "something" to 100 pc and use their full power to crack that "something".
you will have to ask the router if your guess is right or not, and the router computing power is quite limited, and it can be set to answer only one question per second. so yes, it will prevent bruteforcing.

(b) there is no way to limit repeated guesses for an attacker that don't also limit repeated guesses for the legitimate end user. This sounds like a money-making scheme for the router industry to force people to throw their routers away every time they forget their password.

true that it will limit legit users too, but you can allow say 10 times without any delay and after that enable the bruteforcing protection that answer only once per second (or slower).
this is a standard protection that every website uses and you never notice it.
user: one guess every 10 seconds (time needed to write the password)
attacker: 100 guess per second (automated)
while for the forgot password, well you can just reset the router.

654654108July 13, 2018 7:25 AM


Wi-fi password is not used to encrypt your transmisison, but only to connect your device and negotiate encryption between your device and router. You can't use captured trafic to validate passwords you are trying. This was true in WPA2 as well, but WPA3 is doing better job in initial handshake, so you can't get hold of password or encryption key (which is constatly changing) by capturing traffic.

Somewhat less confused nowJuly 13, 2018 8:09 AM

Thanks for that explanation

If I understood you right then really they aren't making "a single guess against that offline data" as previously stated at all, but submitting a password guess to the router then?

Dan HarkinsJuly 13, 2018 1:38 PM

@Some Guy, do I know you? Have we worked together? The things I
take as personal attacks are when people publicly accuse me of
being an NSA plant out to subvert the Internet (which someone did
both in email and in print). So yea, I take personal attacks as
if they're personal attacks.

I'm not sure which "multiple cryptographers" you're talking about.
Dragonfly got lots of comments, some from prominent cryptographers
and some of the comments were security critical. They were all
addressed. Dragonfly has been proven secure in the random oracle
model. It's an unencumbered and proven secure protocol.

It's a balanced PAKE though and that may not be the best thing for
some deployments. It is also susceptible to side channel attack if
the mitigation against it, which is well documented in the
specification, is not employed.

It was adopted by 802.11 for the mesh networking standard because
it's unencumbered and balanced-- it is a true peer-to-peer protocol
which is required for mesh. That was back in 2009 or so. There
has been plenty of time for someone (looking at you @Some Guy) to
specify a different PAKE for 802.11 but somehow no one did and
the "shenanigans" were basically to take the only thing in 802.11
that fixes the well-known flaws of WPA2-PSK and to certify it.
Sorry the story is such a boring let down... no shenanigans.

AlexT July 13, 2018 2:37 PM

One thing that worries me is that this has been worked on without public scrutiny. They "just" came up with those standards... They might be well thought out (time will tell) but I can't even get the full technical details. I'm out is not how good security is achieved.

Dan HarkinsJuly 14, 2018 12:19 PM

@Ross Snider

What do you mean "an NSA backed standard"? Can you please substantiate that statement?

Regarding the weaknesses of dragonfly:
1) there is _one_ known side channel attack if one chooses to do the "hunting-and-pecking" version of PWE assignment. One could, alternatively, use any of the techniques in draft-sullivan-hash-to-curve that are not susceptible to side channel
attack. In any event, there is a mitigation against side channel attack in the
specification anyway so I don't think this is an issue.
2) what are the "several known active attacks"? I'm very curious to learn of these.
3) not quite right. It's a balanced PAKE so the implication is that the credential
is stored in a format that is directly usable by the protocol. Even asymmetric PAKEs
store a password-derived credential in a manner that is brute force-able. That's
just the nature of the PAKE. If you have a better idea please describe it.
4) what malicious parameters are you talking about? Please let me know as this issue
should be addressed if it really is an issue.
5) what is the crypto fragility you are talking about? Validating received elements?
That is prudent for every protocol that does public key crypto. On the other hand,
if there's some issue other than element verification then please do bring it to
my attention!

Can you please answer these questions? If you are talking about actual problems then
you should bring them to the attention of the industry. You know how to get in touch
with me if this forum is not appropriate.

In any event, use a VPN, yes! WPA3 only protects the air, nothing more.

JoaoJuly 16, 2018 9:12 AM

Idea to create a new wireless protocol and create a new certification program... call it "Advanced Wireless Security" or something else.

Simple improvements:
- All devices must have their own public/ private like the “M-511” elliptic curve (or some other, maybe one that already resists Quantum Computing attacks);
- The administrator/ proprietary must be able to change the public/ private keys (in all it’s equipment's);
- Then it can request, optionally, if defined by the administrator/ proprietary, a password to provide also symmetric encryption;
- The public key must be broadcast to the air;
- The public key must be use to make the first encrypted and authenticate communication tunnel, and if any symmetric key (“password”) is also configured then it must create a new encrypted tunnel inside the previous. In both cases it also must create temporarily session keys that change at every new connection and also when it goes by a predetermined time (by default 60 minutes, but configurable).

This way the devices would refuse to connect to devices that don't have the correct private key to the public key, would simply not work and even if the public key is defeated it still uses the symmetric encryption part if it is set. The session keys are used to prevent anyone from recording the data and then able to access the contents after breaking the public key and symmetric key, at least the data recorded until that time should not be easily converted into legible information.

To add the device it would need to be configured: could be by snapping a picture at some QR code; or adding the verification hash of the device it’s trying to connect (derived from the public key); or by accepting the certificate in the "air" after confirming the hash is the same as indicated by the provider of the infra-structure and then entering the password if necessary; or by adding a configuration file with those information's; or by, optionally, pressing a key (when the administrator actives that option, to prevent not authorized persons from adding new devices) or by using the main configuration panel of the device to activate to then exchange the information using light waves (infra-red, laser or other technology) so that it can’t be so easily intercepted and interfered when not in direct line of sight... if the device can't do any of that, it must not be used.

The device that provides the wireless connection to access the network must have the option to allow the administrator/ proprietary to just allow to connect devices that it wants, by adding the hashs of the their public keys.

WiremanJuly 20, 2018 4:28 PM

@realfakenews : "The whole thing needs re-working from scratch, and needs to get rid of all the junk that's included."

All that junk is a problem because the default implementations of most WiFi stacks turn it all on. Recently recompiled my WPA2, and disabled 70% of the default=on options. So, the problem on the hardware side reflects back to the software side. Kind of a mess, which is why I recently bought a bunch of ethernet cable LOL ...

Leave a comment

Allowed HTML: <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre>

Photo of Bruce Schneier by Per Ervland.

Schneier on Security is a personal website. Opinions expressed are not necessarily those of IBM Resilient.